Everything in AI/tech this week in 5 minutes

Happy Sunday. Here's what actually mattered in tech this week, broken down so it's worth your five minutes, then a bit on what's been on my mind. Let's get into it.

Three new frontier models dropped in under 48 hours. OpenAI shipped GPT-6 Sol and Luna, Anthropic put out Claude Opus 5.5 barely an hour later, and Meta and Alibaba were right behind. When four labs launch into the same capability band in the same week, the model stops being the moat and price does. We're now at the point where the frontier is a commodity fight, not a science fair. But it's good news if you build with these things because the best models keep getting cheaper and the gap between "best" and "good enough and free" is basically gone. Read more

Anthropic's Claude found something new in DNA, and even Anthropic isn't sure what it does. They pointed Claude at a huge database of viral DNA with a single prompt, then roughly 950 AI agents ran for 21 hours and flagged a previously undescribed CRISPR-like enzyme system. A CRISPR pioneer at MIT reviewed the pre-print and called it genuinely worth a closer look. The function is still a mystery, but that's almost the point, this is the clearest sign yet that AI isn't just summarizing science, it's starting to surface things humans hadn't found. If you're studying anything technical right now, this is the version of "AI in research" actually worth paying attention to. Read more

OpenAI, Google and Anthropic are teaming up to regulate themselves. The three biggest labs are setting up an independent body (tentatively "SAFA") to define safety standards, vet models before release, and set the rules for reporting incidents, aiming to launch by early 2027. The original plan was a public-private partnership, but that stalled in Washington, so the labs are now building it without the government. Whichever way you read that, either responsible adults filling a gap or the players writing the rules they'll be judged by, it's worth watching who ends up holding the pen. Read more

Google is launching AI chips into space next week. Project Suncatcher puts four of Google's TPUs on a SpaceX Falcon 9 on October 1 to test whether you can run AI on solar power in orbit. For now the chips can only run about 15 minutes before they overheat (yes really), and Google already has two more satellites planned for 2027 with the long-term goal of a network of 80 flying in formation. What it means: AI's real bottleneck now is power and cooling, not chips, and "move the data center to space where the sun never sets" is genuinely one of the plans on the table. Read more

I just spent 3 days with Goodnotes in New Orleans as part of their Goodnotes creator club and I got to meet some of the most amazing creators I’ve met so far. It was a super diverse group of people with creators in all sorts of different niches: medicine, art, philosophy, fashion, tech, entrepreneurship… And everyone has their own unique perspective on content creation and I really loved that. It really made me realize that content creation is not just about filming videos, it really is its own kind of art and everyone approaches it differently. Some people put information and value first, while others put emotion and creativity first and everything finds its right audience. Hearing stories from creators who are doing this full time for 10 YEARS was insane to be honest. I feel quite like an imposter with my humble 1.5 years of experience, so naturally I was super curious and asked loads and loads of questions. All in all it was super cool meeting a lot of the people that I used to take as inspiration a year ago, and it just goes to show how anything is possible so never let people tell you otherwise.

🎬 My latest content:

👥 Community Section:

Have a great week,

Chris